Lyrics Across 38,467 Songs Lost Feeling and Gained Anxiety. The Release Never Says How Words Were Scored.

Willie Nelson has more songs in the dataset than anyone else. Miles Davis, with the fewest lyrics, sits at the other end. Between them are 38,467 songs from 230 artists, collected by Michigan State University researchers and checked by hand by about 30 research assistants, and one part of the pattern ran against what the team predicted.

The study, published October 15 in Psychology of Aesthetics, Creativity, and the Arts, covers lyrics from 1947 to 2024. Lead author Logan Gibson, with Jan Brascamp and William Chopik, tracked two things at once: how each artist's words changed over a career, and how artists from different eras differed from one another. Both views pointed the same way. Lyrics got less emotional and less social.

A career effect that ran the wrong way

The researchers expected musicians to mellow and warm up with age, the way people tend to in psychology studies of aging. They did not. Within individual careers, both positive and negative words declined, and so did mentions of family and friends. Gibson's reading is that lyrics may be "strongly shaped by artistic, cultural or industry-level forces," meaning something bigger than any one songwriter's life.

The era comparison is where the names come in. The Supremes and Marvin Gaye ranked high on positive emotion and social references. Nine Inch Nails and the Sex Pistols ranked low. Older country and blues artists such as Hank Williams and Patsy Cline are overrepresented in the sadness and negative categories, which is a reminder that older music was not simply sunnier.

What the release leaves out

Start with the measuring stick. The university release and the outlets that republished it describe "positive and negative words" and "anxiety-related language" but never name the dictionary or software that sorted the words, and they report no effect sizes. Without those, a reader cannot tell whether anxiety words doubled or crept up a fraction of a percent. I could not reach the paper itself, because the publisher's page returned a 403 error to my request, so this piece runs on the release, a published record of the paper's title and date, and earlier research.

There is also the question of who is in the sample. The release does not say how the 230 artists were chosen, so the result describes those artists and not popular music as a whole. Every outlet I found traces back to the Michigan State release, which means the coverage is one account repeated, not several accounts agreeing. And word counts are not feelings. A song can be devastating without using a single sadness word, and a lyric sheet says nothing about how a track sounds.

Not the first lyrics study to find a drift

An older study gives the new one some context. In 2011, a University of Kentucky team led by Nathan DeWall analyzed lyrics from Billboard top-10 songs between 1980 and 2007 and reported more first-person singular words such as I and me, fewer first-person plural words such as we, and more narcissistic language over time. DeWall summed it up this way: "Over time, songs are becoming more about 'me' and less about 'us.'"

The designs differ. DeWall's sample was chart hits across 28 years, and the Michigan State one follows discographies across roughly eight decades. That the two point the same way on social language is a point in the finding's favor. It is not proof, since both rest on word counts and neither measures what listeners feel.

“Music is not just entertainment.”

The open question is the one the researchers themselves raise. If lyrics are drifting because of industry or cultural pressure, something in how songs get made and chosen is changing, and a word count alone cannot say what. The full paper, with its methods and effect sizes, is where that answer would be.